If you've been watching your Claude Code token meter spin faster than expected, you're not alone—and the community has a pretty clear theory about why. A recent discussion on DEV.to highlights what many developers have quietly grumbled about: Claude Code talks too much. With over 375,000 installs, it's clear Anthropic built something useful—but that usefulness comes with a verbosity tax that can eat through your usage limits quicker than you'd like.

The Verbosity Problem

According to the analysis, burning through a Claude Code session isn't about tackling hard problems—it's about the agent's narration pattern. For every single task, users report watching this cycle unfold: first it announces its plan, then it restates your context (sometimes redundantly), explains the diff it's about to write, actually writes it, and finally summarizes what it just did. Multiply that by dozens of tasks in a typical coding session and you've got serious token consumption before lunch.

Why This Matters for Developers

The frustration isn't that Claude Code is unhelpful—it's that the constant narration interrupts flow state. Developers want their AI assistant to execute, not perform a running commentary. When you're doing repetitive refactoring or batch file edits, hearing "I'm going to update this function now" followed by "I've updated this function" adds zero value while adding significant token cost. This becomes especially painful when working on larger projects where you might run dozens of Claude Code commands in a single session. The verbose output that seems charming during a quick five-minute task becomes exhausting (and expensive) over hours of development work.

What Developers Are Saying

The DEV.to post struck a nerve, resonating with developers who've noticed their usage charts climbing faster than their code quality improved. The 375,000 install figure demonstrates this isn't a fringe complaint—it's a widespread experience affecting a significant chunk of the developer community using Anthropic's CLI tool.

Key Takeaways

  • Claude Code's constant narration (plan → context restatement → diff explanation → write → summary) burns tokens fast
  • 375,000 installs indicate this is a mainstream complaint, not an edge case
  • Verbosity interrupts deep work and flow state during longer coding sessions
  • The feature set is valuable enough that people keep using it despite the verbosity issue

The Bottom Line

Anthropic built Claude Code to be transparent about its actions—and for learning or exploration, that's genuinely useful. But for seasoned developers grinding through routine tasks? That transparency becomes noise. Someone needs to ship a "silent mode" flag before this tool loses the power users it's trying to keep.